Paper on PaLM

Compute

854 H100e

Chips

6,144

First operational

Apr 5, 2022

Paper on PaLM is a large GPU cluster operated by Google, located in the United States, in the private sector. It came online around April 2022. At that moment it ranked number 15 among known clusters.

It is built on 6,144 Google TPU v4. That amounts to about 854 H100e of AI compute. The cluster draws roughly 4.45 MW. Its hardware is estimated at $56M.

Epoch AI rates the certainty of this record as confirmed.

Full record

Owner: Google
Users: Google
Sector: Private
Status: Existing
Country: United States
Primary chip: Google TPU v4
Primary chip count: 6,144
Total AI chips: 6,144
H100 equivalents: 854
8-bit throughput: 1.7×10¹⁸ OP/s
Power capacity: 4.45 MW
Hardware cost: $56M
First operational: Apr 5, 2022
Rank when it came online: 15
Certainty: Confirmed